Intelligent CPE equipment management system based on visual identification

Abstract

The invention is suitable for the technical field of communication network equipment management, and particularly relates to an intelligent CPE equipment management system based on visual identification, and the system comprises a preprocessing module which is used for collecting equipment image information and user image information; the equipment identification and authentication module is used for identifying the equipment image information and the user image information by adopting a small sample target identification algorithm, and identifying a user and equipment; and the intelligent device matching module is used for adjusting the network configuration strategy of each device according to the physical position of the device and the activity scene of the user, and providing a higher bandwidth for the device with a higher priority. A camera is used for capturing equipment and user images, equipment types and user identities are recognized through a visual model, known equipment is automatically authenticated, and network configuration and priorities of the equipment are automatically adjusted by recognizing physical positions and user activities of the equipment; unauthorized equipment is prevented from accessing the network, equipment authentication is prompted to the user, and the network security is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of communication network equipment management, and in particular relates to an intelligent CPE equipment management system based on visual recognition. Background Art

[0002] As the complexity of communication networks, home or enterprise networks increases, the demand for managing access devices is also growing. Existing CPE devices generally manage access devices through physical or logical methods, mainly relying on manual configuration or static authentication based on MAC addresses. Among them, physical management determines which devices can access the network through physical ports. This method usually requires manual configuration of CPE devices, through CLI (command line interface) or Web management interface, to specify that certain ports or interfaces can only connect to specific types of devices. Logical management usually involves configuring VLANs, subnets or other network policies to manage access devices. This process can be achieved through the configuration of network devices such as routers or switches. The above two methods mostly rely on manual configuration by administrators, and generally choose to use static authentication based on MAC addresses. This is a common device authentication method based on MAC address filtering. In this method, the CPE device will save a table to record the MAC addresses of devices allowed to access. If the MAC address of a device is not in the allowed list, it will not be able to access the network. This method is called static authentication because the MAC address is manually configured or statically assigned, and the device will not change dynamically after accessing the network.

[0003] CPE devices generally manage access devices through physical or logical means. The existing shortcomings mainly include complex manual configuration, easy forgery of MAC addresses, lack of dynamicity, etc. Manual configuration of devices requires administrator intervention, which is labor-intensive and prone to errors. MAC addresses can be forged, so static authentication based on MAC addresses is not a very secure authentication method. Static authentication methods lack dynamic adaptability, cannot dynamically adjust network access policies according to device characteristics and user behavior, and cannot optimize network configuration according to device usage. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent CPE device management system based on visual recognition, aiming to solve the problems that the static authentication method lacks dynamic adaptability, cannot dynamically adjust the network access strategy according to device characteristics and user behavior, and cannot perform better network configuration optimization according to device usage.

[0005] The present invention is implemented as follows: an intelligent CPE device management system based on visual recognition, the system comprising:

[0006] A preprocessing module, used to collect device image information and user image information;

[0007] The device identification and authentication module is used to identify the device image information and the user image information using a small sample target recognition algorithm, and identify the user and the device;

[0008] The intelligent device matching module is used to adjust the network configuration strategy of each device according to the physical location of the device and the user's activity scenario, providing higher bandwidth for devices with higher priority.

[0009] Preferably, the preprocessing module stores user information and device information, and one user can be bound to multiple devices, and one device can be bound to multiple users.

[0010] Preferably, the small sample target recognition algorithm adopts a lightweight convolutional neural network: the first layer is the input layer, which is used to input images; the second layer is the convolution layer, which is used to extract image features; the third layer is the pooling layer; the fourth layer is the convolution layer, which uses twice the convolution kernel of the second layer, and the activation function adopts the Hard Swish function; the fifth layer is the pooling layer, which uses the max-pooling method for sampling; the sixth layer is the dropout layer, which is used to set the parameters of some neural nodes to 0 according to the set probability; the seventh layer is the flattening layer, which is used to fold the spatial dimension of the input into the channel dimension to connect to the complete connection layer; the eighth layer is the fully connected layer, and the activation function adopts the Hard Swish function; the ninth layer is the output layer, and the activation function adopts the SoftMax function.

[0011] Preferably, in the small sample target recognition algorithm, the second layer is a convolutional layer, and the activation function uses a Hard Swish function.

[0012] Preferably, the third layer is a pooling layer, which uses a max-pooling method for sampling to reduce the dimension of the data and compress the characteristics of the image.

[0013] Preferably, the sixth layer is a dropout layer, and the probability of setting is set to 0.02 to reduce the probability of overfitting during training.

[0014] Preferably, the user's activity scenes include at least home, company, shopping mall, bedroom, living room, office, conference room and store.

[0015] Preferably, when the smart device matching module detects that a user is performing a specific activity in a specific activity scene, a higher network priority is provided for the corresponding device used by the user, and the occupancy of network bandwidth by other devices with lower network priorities is reduced.

[0016] Preferably, when a user uses a device, traffic analysis, network protocol, and visual recognition are used to determine the usage status of the device. During the traffic analysis, the process that the device is running is determined based on the data packet size, transmission speed, and transmission time.

[0017] Preferably, in the process of determining the device usage status based on the network protocol, when a WebRTC or Skype protocol used for video calls is detected, it is determined that the user is using the device for communication.

[0018] The present invention provides an intelligent CPE device management system based on visual recognition, which uses visual recognition technology to realize intelligent device management and user recognition in CPE devices, uses a camera to capture device and user images, identifies device types and user identities through visual models, automatically authenticates known devices, and automatically adjusts the network configuration and priority of the device by identifying the physical location of the device and user activities; prevents unauthorized devices from accessing the network, prompts users to perform device authentication, and ensures network security. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram of a process flow of a smart CPE device management system based on visual recognition provided by an embodiment of the present invention;

[0020] Figure 2 An architecture diagram of a small sample target recognition algorithm provided by an embodiment of the present invention;

[0021] Figure 3 A schematic diagram of the overall process framework of the intelligent CPE device management system based on visual recognition provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0023] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0024] like Figure 1 , Figure 2 and Figure 3As shown, an intelligent CPE device management system based on visual recognition provided by an embodiment of the present invention includes:

[0025] The preprocessing module is used to collect device image information and user image information.

[0026] In this module, the CPE device management system is initialized before use, and the device type and user image information that need to be connected are collected. One user can bind multiple types of devices, and the same device can also bind multiple users, and the device network is configured according to different user situations. The visual recognition model and device information storage obtained through the preprocessing module pave the way for subsequent use.

[0027] The device identification and authentication module is used to identify device image information and user image information using a small sample target recognition algorithm, and to identify users and devices.

[0028] In this module, the camera integrated in the CPE device is used to capture device and user image information in real time, and the visual recognition model is used to identify the device type and user identity. The system uses a small sample target recognition algorithm based on a lightweight convolutional neural network to process and analyze the captured images, identify different devices such as smartphones, tablets, smart TVs, and confirm the user identity through facial recognition technology. The small sample target recognition algorithm based on a lightweight convolutional neural network can achieve good target recognition effects under limited computing resources and a small number of training samples by introducing a lightweight target recognition network structure. The core of this type of algorithm is to reduce the computational complexity and parameter amount of the network, and at the same time use various small sample learning strategies to deal with the problem of insufficient data, thereby achieving efficient and accurate target detection effects under limited conditions. Different from traditional CNNs that usually use 3 or more convolution blocks with too many training parameters, the present invention proposes a small sample target recognition algorithm based on LWCNN, which has 2 convolution blocks and a small number of convolution kernels. Its network structure is as follows: Figure 3As shown, the first layer is the input layer, input image; the second layer is the convolution layer, the activation function uses the Hard Swish function, this layer is used to extract image features; the third layer is the pooling layer, which uses the max-pooling method for sampling, this layer can reduce the dimension of the data and compress the characteristics of the image; the fourth layer is the convolution layer, which uses twice the convolution kernel of the second layer, and the activation function uses the Hard Swish function, which can further extract image features; the fifth layer is the pooling layer, which uses the max-pooling method for sampling, and this layer further compresses image features; the sixth layer is the dropout layer, which sets the parameters of some neural nodes to 0 according to the set probability, and the probability is set to 0.02 in the present invention to reduce overfitting during training; the seventh layer is the flat layer, which folds the input spatial dimension into the channel dimension so as to connect to the complete connection layer; the eighth layer is the fully connected layer (dense layer), and the activation function uses the Hard Swish function; the ninth layer is the output layer, and the activation function uses the SoftMax function. Finally, it is identified that the user and the bound device are the user images and device information collected and stored during initialization, and the network connection is performed for the device.

[0029] Based on the information stored in the pre-processing module, the system can automatically authenticate known devices, avoiding the tedious process of repeated identity authentication. For unauthorized devices, the system will automatically block them from accessing the network and prompt the user to add a new device authentication to ensure the security of the network. For security reasons, the system can record the identification and authentication history of all devices for administrators to audit and monitor.

[0030] The intelligent device matching module is used to adjust the network configuration strategy of each device according to the physical location of the device and the user's activity scenario, providing higher bandwidth for devices with higher priority.

[0031] In this module, the system automatically adjusts the network configuration and priority of the device according to the physical location of the device and the user's activity scenario. The activity scenario includes but is not limited to macro environments such as home, work, and shopping malls, and can also be subdivided into micro environments such as bedrooms, living rooms, offices, conference rooms, and shops. In a home environment, when the system recognizes that the user is using a streaming service to watch high-definition videos in the bedroom, the system will automatically increase the network priority of the device in the bedroom to ensure that the video can be played smoothly and reduce the network bandwidth usage of other non-critical devices. In the conference room of a unit or shopping mall, when the system detects that a video conference is in progress, it will also give priority to the network bandwidth of conference-related devices to ensure the smooth progress of the meeting. In terms of priority, application scenarios involving real-time communications such as video conferencing, online games, and high-definition streaming media playback that require high bandwidth and low latency will be given higher priority. Some non-real-time application scenarios that do not require high bandwidth, such as web browsing and file downloading, will be assigned lower priorities. The specific priority settings can also be customized according to the actual needs of the user, aiming to improve the utilization efficiency of network resources and optimize the user experience.

[0032] Provide personalized network environment settings for different users, such as common bandwidth priority and other different needs. The system can perceive different activity scenarios such as user's home entertainment, office mode, and automatically adjust the network configuration of the device according to the scenario. CPE equipment identifies the user's device usage status through traffic analysis, network protocol and visual recognition. By analyzing the characteristics of traffic packet size, transmission speed, transmission time, etc., CPE equipment can identify which devices are making high-definition video calls or running high-traffic consumption software; at the same time, it can also analyze network protocols to identify which devices are using specific protocols for communication, such as WebRTC, Skype protocol and other video call protocols for video calls or application protocols such as high-traffic consumption video streaming protocols; it can also be recognized by vision. Although visual technology is mainly used for intelligent matching and identification of location scenarios, in some cases, it can also assist in identifying the device's usage status. By capturing the video call interface on the device screen through the camera, it can be indirectly judged that the device is making a video call. The above recognition mechanisms do not exist in isolation, but complement and work together. Therefore, when one of the users is making a high-definition video call and needs to ensure the network, the system will automatically allocate more bandwidth to the relevant devices to ensure the smoothness of the call.

[0033] Compared with traditional CPE device management, the present invention uses vision-based intelligent CPE device management, improves device management security through intelligent identification and authentication modules and intelligent device matching modules, dynamically identifies access devices, prevents unauthorized devices from accessing the network to prevent information leakage, and can automatically adjust the network configuration and priority of the device by identifying the physical location of the device, providing a better network environment experience. At the same time, a small sample target recognition algorithm based on a lightweight neural network is used, which can still achieve good target recognition effects under limited computing resources and small training samples.

[0034] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent CPE device management system based on visual recognition, characterized in that: The system comprises: A preprocessing module, used to collect device image information and user image information; The device identification and authentication module is used to identify the device image information and the user image information using a small sample target recognition algorithm, and identify the user and the device; The intelligent device matching module is used to adjust the network configuration strategy of each device according to the physical location of the device and the user's activity scenario, providing higher bandwidth for devices with higher priority.

2. The intelligent CPE device management system based on visual recognition according to claim 1 is characterized in that: The preprocessing module stores user information and device information. One user can be bound to multiple devices, and one device can be bound to multiple users.

3. The intelligent CPE device management system based on visual recognition according to claim 1 is characterized in that: The small sample target recognition algorithm adopts a lightweight convolutional neural network: the first layer is the input layer, which is used to input the image; the second layer is the convolution layer, which is used to extract image features; the third layer is the pooling layer; the fourth layer is the convolution layer, which uses twice the convolution kernel of the second layer, and the activation function adopts the Hard Swish function; the fifth layer is the pooling layer, which uses the max-pooling method for sampling; the sixth layer is the dropout layer, which is used to set the parameters of some neural nodes to 0 according to the set probability; the seventh layer is the flattening layer, which is used to fold the spatial dimension of the input into the channel dimension to connect to the complete connection layer; The eighth layer is a fully connected layer, and the activation function uses the Hard Swish function; the ninth layer is the output layer, and the activation function uses the SoftMax function.

4. The intelligent CPE device management system based on visual recognition according to claim 3 is characterized in that: In the small sample target recognition algorithm, the second layer is the convolution layer, and the activation function uses the Hard Swish function.

5. The intelligent CPE device management system based on visual recognition according to claim 3 is characterized in that: The third layer is the pooling layer, which uses the max-pooling method for sampling to reduce the dimension of the data and compress the characteristics of the image.

6. The intelligent CPE device management system based on visual recognition according to claim 3 is characterized in that: The sixth layer is the dropout layer, and the probability is set to 0.02 to reduce the probability of overfitting during training.

7. The intelligent CPE device management system based on visual recognition according to claim 1 is characterized in that: The user's activity scenarios include at least home, company, shopping mall, bedroom, living room, office, conference room and store.

8. The intelligent CPE device management system based on visual recognition according to claim 7 is characterized in that: When the smart device matching module detects that a user is performing a specific activity in a specific activity scene, it provides a higher network priority for the corresponding device used by the user and reduces the network bandwidth occupied by other devices with lower network priorities.

9. The intelligent CPE device management system based on visual recognition according to claim 8, characterized in that: When a user uses a device, traffic analysis, network protocols, and visual recognition are used to determine the device's usage status. During the traffic analysis process, the device's running process is determined based on the data packet size, transmission speed, and transmission time.

10. The intelligent CPE device management system based on visual recognition according to claim 9, characterized in that: In the process of determining the device usage status based on the network protocol, when a WebRTC or Skype protocol used for video calls is detected, it is determined that the user is using the device for communication.

Images